Document 11049611

advertisement
Dewey
1
MASS. »NST. TFCH.
JUN 11
I
1971
j
DEWEY LIBRARY
The Management Process, Management Information
and Control Systems, and Cybernetics
by
Zenon S. Zannetos and Jarrod W. Wilcox
CO^I
412-69
August 1969
'
RECEIVED
JUN 23
•
'•
T.
1971
LIBRARIES
THE MANAGEMENT PROCESS, MANAGEMENT INFORMATION
AND CONTROL SYSTEMS, AND CYBERNETICS
Zenon
S.
Zannetos and Jarrod W. Wilcox
The purpose of this paper is to examine the impact of cybernetics
on management of business organizations.
If we define "cybernetics
as
the science of control through communication, using the conceptual frame-
work constructed by Wiener and modified by others,
then we can definitely
find ample examples where cybernetic principles exert significant influ-
ence on management activities.
Perhaps one may in the broadest sense
consider cybernetics as a subset of management or possibly equate the
two.
For reasons which will become clearer later, we prefer to view
cybernetics as a subset of management.
Despite this more or less definitional harmony, there is, however, a
legitimate question as to what extent we can attribute such uses of cyber-
netic principles to the explicit emergence of this field of scientific
pursuit.
We can find evidences in the areas of managerial economics and
managerial control where similar principles have been used implicitly for
years, although typically not consciously.
However, this does not detract
The framework puts emphasis on the information flows through the
cybernetic system and on the way knowledge is utilized by such systems
to gain their ends in a self-correcting manner.
Both of the Sloan School of Management of M.I.T., Cambridge, Mass.
,
5-^7445
2.
from the ultimate influence of cybernetics on management; as in many
other areas of human pursuit, it is not so much the invention of new
theories and principles which leaves its indelible mark on progress but
rather the articulation and internalization of notions which previously
have had only an unconscious empirical existance.
In order to develop our arguments we will
(i)
start with a con-
ceputal framework which will enable us to link management of business
organizations and cybernetics,
(ii)
indicate a conceptual representation
of the managerial process and identify its major components,
(iii)
assess the impact of cybernetics on management across some of the most
essential elements of managerial activities, and finally (iv) present
some concluding remarks.
I.
A Conceptual Framework for Relating Cybernetics and Management
In attempting to develop our framework we wish to distinguish
between four kinds of entities:
1.
2.
3.
4.
Computational Machines
Cybernetic Systems
Organisms
Organizations
We will call devices which transform, after a time delay, sequences of
changes in the states of input channels to sequences of changes in the
states of output channels computational machines
2
3
2
or transducers.
3
See Marvin Minsky, Computation, Finite and Infinite Machines
See Ross Ashby
,
An Introduction to Cybernetics
3.
One may usefully categorize such machines by the degree of fixity
of state possibilities (finite machines versus Turing machine equivalents),
by the degree of invariance in the relationship between input and output
sequences (deterministic versus probabilistic)
,
and by the degree of
discreteness (discrete versus continuous)
The partially overlapping theories of automata,
transmission,
and cybernetics
and design optimalities
In general,
'
information
treat the various kinds of behaviors
of certain subsets of computational machines.
the more mathematically developed portions of these
theories deal with rather abstract kinds of computational machines which,
although internally consistent, do not allow us to get close enough to
the detail of reality necessary for our purposes.
We need a theory of
computational machines physically embedded in an environment, where
the combined system of machine and environment is subject to physical
laws, particularly the law of non-decreasing entropy.
When a computational
machine through a flow of information is able in a self-correcting manner
to gain and maintain for an appreciable time internal states which are
not consistent with maximum entropy in such a reality-oriented environ-
ment, we will call it a "cybernetic system."
When it can do so for an
indefinite period in a non-artificial environment we will call it an
"organism."
This notion seems useful even though what we mean by
"indefinite" and "non-artificial" is
4
as yet imprecise.
See von Neumann, The Computer and the Brain
See Minsky, op.cit.
See Shannon, The Mathematical Theory of Communication
See Wiener, Cybernetics
When such organisms couple together to form a conglomerate
cybernetic system but retain an ability to decouple themselves and
resume their independence we will call the conglomerates "organiza-
tions."
Organizations may or may not behave as organisms.
While it is clear that all organizations involve organisms and
in turn cybernetic systems and finally computational machines, organi-
zations also may have properties not considered by the more basic
theories.
Thus one can expect to find applicability of cybernetics to
management necessary, but such applicability may not be sufficient to
either describe fully or "rationalize" the management process.
In
moving from computational machines to cybernetic systems we first necessarily encounter error, uncertainty, and environmental change.
In
moving to organisms we encounter a variety of so-called "irrational" behaviors engendered by limited capacities as opposed to an unlimited environment and bv lack of unitary objectives. Finally, in moving to organizations
we move to phenomena of culture, the gamut of social behaviors and interactive modes, and typically also encounter difficulties in reconciling
conscious and unconscious interdependencies
Accordingly, we must be very careful in relating the findings of
cybernetics to the management of real-world organizations.
II.
The Management Process
Managerial activity in business organizations includes both planning
and control aspects.
Specifically, it has been viewed as including:
1.
Environmental analysis in order to assess the state of
nature regarding both the extern al as well as the
internal environment with which the organization is and
will be dealing.
2.
Determination of go als or alternative directions.
Environ-
mental analysis can provide some "vague" picture regarding
the existing patterns of relationships.
Through the appli-
cation of some "normative criteria," a comparison between
what appears to be and what should be is made, and through
this comparison goals are derived for guiding the particular
entity or organization.
The word "vague" is used above
because often it is not at all clear as to what is being
exactly observed.
And if what is observed is not very clear,
obviously there can be no certainty that the criteria applied
and the comparisons made are the ones most appropriate.
There-
fore, a process of iteration and sequential adaptation may
have to take place before goals are derived.
3.
Transformation of the goals into operationally meaningful
proposit ions, and objectives for action.
In other words,
here are defined the strategies by means of which the goals
and the more concrete objectives are to be achieved through
operations (means to end chains).
4.
Choice of the critical decisions and operations which are
necessary for reaching the objectives and design of an
6.
operational plan for execution
.
Part of the design of the plan
involves planning of the organization structure.
In addition,
this step includes forecasting, definition of resource develop-
ment activities, logistics, and budgeting as well as scheduling
of the operations.
5.
Choice of the signals or indicators which are necessary to
inform management as to whether
t he
execution of operations
is according to the chosen plan and whether the chos en critical
decisions and operations are in effect leading the organization
to the accomplish ment of its objectives
6.
.
Choice of the measureme nts which when applied to the results
from operations will generate
previous step.
t he
signals mentioned in the
The choice of the dimensions across which
measurements are applied as well as the refinement of the
measurements themselves and the size of deviation which will
necessitate replanning and remedial action are determined by
the sensitivity of organizational goals to the environment
parameters, operations, and decisions being monitored.
7.
Ch oice of the management information and co ntrol system which^
will apply the measurements to
generate the chosen signals.
As will be noticed, the last three steps in the management process
are mainly, though not exclusively, addressed to the informational output
of the implementation of managerial plans, and as such refer to the
control process
.
The iterative and sequential nature of managerial
planning and control is manifested in each of the above-mentioned steps.
In practice,
the process is also hierarchical, not only in its totality
but also within each of the above major subactivities.
The above description indicates that the management process has
the essential elements of a cybernetic control system.
In fact, we
have in the above seven steps a whole hierarchy of cybernetic processes.
In order to simplify our exposition we will concentrate on the following
three main elements of the management process:
A.
Goal formation
B.
Hierarchical structuring
C.
Control
The question which we would like to ask is to what extent the past
developments in cybernetics have helped management in each of the above
key elements of the management process.
In so doing, we wish to distin-
guish those areas of management where cybernetic principles and propositions have been applied in an operating sense from those where only
philosophical benefit may have been derived.
This distinction does not
imply that theoretical constructs cannot be used to derive useful opera-
tional consequences or objectively varifiable hypotheses.
Care must be
exercised, however, so that we do not let our methodological purity lead
us to models which are correct and viable within the limits of their
assumptions but yet do not satisfactorily express reality.
Richardson, "Imperfect Knowledge and Economic Efficiency," p. 156.
III.
The Impact of Cybernetics on ManaRement
A.
Goal Formation
If we look at the processes of goal formation, or its relatives
criteria selection, generation of alternatives, and the assessment of
the state of a complex environment, expecially when the latter is
shifting, we will find that cybernetics has had little practical influence on management.
The same can be said, however, for almost any other
field of scientific endeavor with the partial exception of economics.
We have all heard predictions about cybernetic systems which would
model themselves as well as the environment with which these interact
and also model their perception of the process of interaction and model
formation itself (for sequential adaptation)
yet with us.
,
but such progress is not
The reasons for this, of course, is that we do not at this
time understand all the complexities of the process of modeling and goal
formation.
In terms of managerial reality, many management problems and thus
management objectives are imposed upon managers by the economic reality
of markets.
The role of the latter in managerial goal formation has
long been recognized.
As Richardson pointed out, the invisible hand of
Adam Smith is nothing less than the homeostatic control of society; "it
acts as a method of communication, and by integrating the knowledge which
is dispersed in many minds, enables society to work as a whole.
9
Richardson, op.cit., p. 140.
9
9.
Under the perfectly competitive model of entrepreneurial behavior
implied in Adam Smith's framework, equilibrating prices are established
through the unconscious interactions of the markets and within such
markets.
These prices serve under the hypothesized conditions as the
only information necessary to motivate the entrepreneur to allocate
resources efficiently (profit maximizing).
One need not examine the
"internal structure" of the entrepreneurial organization or the decision
making processes of the entrepreneur because the market enforces universal rationality.
But the perfectly competitive model, under which
the market served as the cybernetic control system, put emphasis on
the result of the process, not the process itself or method of adjustment
to disturbances.
Because of this emphasis the theory of managerial
planning and control has been neglected until recently; so also has been
the study of individuals and organizations as information processing
entities.
If we step back and look at the processes of adjustment, of neces-
sity we must examine the relationship of the individual or the organization to the outside world.
Then we find that we must develop models
and hypotheses which state explicitly the critical causal connections,
of paramount significance among which will be hypotheses about how the
individual and the organization learn from experience.
Underlying the
main hypotheses there are others on how perception occurs, on the type
See von Hayek, Simon, March & Simon, Cyert and March among others.
See von Hayek, "Economics and Knowledge."
10.
of filtering which is characteristic of certain organizations and
individuals, on the way problems are structured and decomposed,
and on the process of association, selection and use of incomplete
information in solving the perceived problems.
Even when enough is known to predict the adjustment behavior of
individuals, this does not imply that we can necessarily predict the
behavior of the organization of which they are components; the reverse
is also true.
Organizations like the perfect competitive markets reach
their equilibrium somehow by the use of the bits and pieces of infor-
mation which are processed by individuals.
Again, we have learned that
knowledge of how the individual adjusts and reaches his goals is not
enough to provide a prediction of how the organization behaves under
similar situations unless we know also how these information processors
are inter-related.
This gradual emergence of our understanding of the importance of
viewing economic and business activity in terms of dynamic organizational
adjustment under conditions of uncertainty is no doubt partly a result
of the emergence of the field of cybernetics.
The necessity of understanding the process of adjustment of orga-
nizations is even more critical when we are dealing with oligopolistic
and monopolistic situations.
Characteristic of the latter is the degree
of conscious interdependence among organizations.
In managerial planning,
therefore, no longer we can afford to consider competitive behavior as
part of an uncontrollable given environment to which the organization
11.
reacts.
To be effective, managerial planning must include sophisticated
environmental appraisal and prediction of the competitors' reaction to
an initial action by the organization.
Cybernetic notions have provided
an important philosophical stimulation in the efforts of management
scientists who are attempting to derive organizational goals in situations
characterized by oligopolistic and monopolistic competition.
Essential
elements of the transactions of an organization with its environment can
be modeled in terms of higher order cybernetic control sytems.
Thus in the area of goal formation and related processes, cybernetics
has made some philosophical contribution to management.
In our estimation
the impact of cybernetics would have been much greater if it were not for
certain historical accidents.
The history of cybernetics as a field has been marked by a core of
careful, serious science surrounded by a colorful penumbra of speculation
and loose analogies.
Within this penumbra a number of cultural values
have grown up which seem to us to have no solid base in either theory
or practice.
1.
Two of these relate to the process of goal formation.
Confusion about the relationship between entropy, information, and
cybernetic control.
Many writers have pointed out the apparent connection between the
notions of information and negative entropy.
Properly understood,
12
tV
relationship indicates a basis for a comprehensive theory of organisms.
^See Brillouin, Scientific Uncertainty and Information
12.
Improperly understood, "entropy" is a worse than useless concept which
obscures practical application of the cybernetic approach.
the following quoted remarks.
Consider
13
Suppose that the two sub-systems both begin with the same
Sub-system A uses up a lot of this energy in
amount of energy.
Sub-system B uses
the process of organizing itself internally.
up less of its energy in the process of organizing itself to a
lesser extent.
So A is more organized, and more depleted of
Accordingly, since interaction occurs, energy
energy, than B.
It is now, as
must, by the rules of entropy, flow from B to A.
it were, too late for B to catch up in degree of organization
with A.
It has a decreasing supply of energy available to use for
So the more
organizing itself, while A has an increasing supply.
Eventually A will
organized A feeds on the less organized B.
destroy B altogether (in an isolated system). Note that the
boundary of A, which is its interface with B, has to be visualized
That is, the degree of organizaas advancing into B's territory.
tion moves against the direction of the flow of energy.
Now the business of self-organization becomes finally clear
when it is realized that a system has to be recognized as being
organized when in its most probable state. A prime example of
A seed has to be considered
this occurs in the process of growth.
as a variety-amplifier, for it carries the specification of
But it also carries a temporallysomething larger than itself.
based plan for growing to maturity: a self-organizing capability.
Not only does this plan specify a set of architectural relationThat is, any
ships, it specifies a criterion of maturity.
organic seminal programme that inaugurates and controls growth
This capability applies not only to the
'knows when to stop.'
macrostructure, so that you and I are roughly the right size to
be recognizably human; it also applies to the infrastructure of
the organism: every limb, every organ, every parcel of tissue
however delineated, from cranium to toe-nail, grows to a limit.
During growth, further growth is by definition due to follow;
development (except by massive intervention from outside) cannot
To this extent, a partlybe arrested until the plan is complete.
grown organism is in an improbable state, and is driving towards
Growth can be regarded,
its most probable state adult completion.
The growth process stops
that is to say, as an entropic process.
having been finally
actuality
when the genetic information is used up,
Any form of the entropy equation
and in sum exchanged for potentiality.
Growth, then, is a self-orgawill serve to formalize this process.
nizing activity of a system in which that system 'learns to be what it
—
13
Stafford Beer, Decision and Control The Meaning of Operational
Research and Management Cybernetics p. 347; p. 360-61.
:
,
13.
From the point of view of business organization, conclusions that
there is a simple transformation between energy and structure, that
organized states are most probable, and that a maturing organism represents a mere unfolding of the structure inherent in its "seed," are not
only of limited usefulness but also discouraging.
Clearly, organized
states are most probable, given self-correcting and self-organizing
properties of the organism which are inherently biased toward such states,
but this is a probability itself conditional on an unlikely event, the
existence of these properties.
Without them, the most probable state of
nature in business management, is stagnant mediocrity and ultimately
bankruptcy.
In the area of management goals cannot be considered as fixed.
Innovation is the life-line of business organizations, although such
innovations may initially increase the entropy in the total entity.
If we were to accept fixity of goals as a fundamental premise of cyber-
netics, we would then have to relegate cybernetic influence only to the
control of operations.
Operational control systems are homeostatic but
we hope that over time the goals change.
Maturation of organizations
consists partly of the incorporation of negative entropy extracted from
the environment rather than merely the disposition of existing resources.
A more useful and correct notion for management is to regard "know-
ledge," or more broadly speaking, "resources," as that which enables a
cybernetic system to reduce its own entropy, "information" as additions to
knowledge, and "cybernetic control" as the process by which a cybernetic
system utilizes this knowledge.
14.
2.
Over-reliance on random, as opposed to selective, self-organization.
A popular notion in the field of cybernetics is that of "requisite
variety."
14
A cybernetic system seeking to control disturbances of
great variety must itself be of great (requisite) variety.
One way this
variety can be generated is through random mutation and search.
Such
procedures are, however, inefficient as compared to those which can utilize
even very limited knowledge of the search space.
We have come to appreciate the limitations of such purely random
adaptation through sober experience with the "Perceptron" and similar
efforts.
If we wish to develop the capacity to cope with very high-
variety disturbances, and to accomplish it within a time reasonable for
the necessary action, we must use selective search or adaptation.
Stafford
Beer has proposed setting fast machines to work trying at random until
they get acceptable solutions as a practical means of controlling complex
industrial processes.
Such purely random control is likely to find
limited application in industry not only because of the urgency of decisions
but also because in many cases goal-oriented search can yield simple heuris-
Even though Beer proposes
tics for "pruning" the alternative action trees.
a two-part machine one of which vetoes
(selects) among random proposals
of the other, he nevertheless still faces the basic problems of explosive
alternatives.
True, he has saved the human having to explore many of the
random proposals, but the machine still must go through an unselected set
14
Ross Ashby, Introduction to Cybernetics
,
p.
206, p.
245.
Marvin Minsky, "Steps toward Artificial Intelligence," in Computers
and Thought Feigenbaum and Feldman.
,
Stafford Beer, Cybernetics and Management
,
p.
136-148.
15.
of possibilities.
A fact that is often overlooked is that the computer,
in spite of its speed, may still find this set too large to handle within
a
practical time span.
For example, it has been estimated that a random
search for the best opening move in the game of checkers might entail
checking on the order of perhaps 10
40
move sequences, which, at three
choices per millisecond, would still take 10
21
centuries to consider.
17
These astonishing figures suggest that even for somewhat simpler tasks,
it would be wise for the designer of such a machine to build in as much
knowledge as practicable to be utilized not just to veto random proposals,
but to guide in the generation of non-random proposals.
We do not suggest
complete elimination of randomness even if it could be achieved; we
strongly believe that a degree of randomness is useful for challenging
established thought processes and solutions.
But this is a good example
of the need to carefully check the simplified assumptions on which a
theoretical result is based against the demands of practical application.
B.
Hierarchical Structuring
One of the notions which has had a tremendous impact on the practice
of management is that of an organizational hierarchy with its concomitant
substabilities.
At first the progress through the utilization of these
notions has been unconscious, hence passively accepted and evolutionary,
l^ien,
however, organizations grew out of the small-scale entrepreneurial
A. L. Samuel, "Some Studies in Machine Learning Using the Game
of Checkers," in Computers and Thought , Feigenbaum and Feldman (Eds.)
16.
fold to the degree of complexity represented by large-scale oligopolistic
and monopolistic environments, crises enforced a re-examination of both
the process of evolution and the principles embodied in the various orga-
nizational configurations.
of both practitioners
It was out of such soul searching on the part
and "students" of management that the role of
hierarchical structures with articulated substabilities started to emerge.
Tlie
soul searching and the need for progress are by no means over.
It is
our belief, however, that a strong foundation for sustained progress has
been set once the profession has realized that organizational structures
can be included
and used as an active managerial tool as well as resource.
So no longer do we leave the design of organization structures to crises
which emanate out of the transactions between the organization and its
environment, nor do we accept that there exists an immutable natural homeostatic state to which organizations must be allowed to gravitate.
This
does not deny, however, the usefulness of homeostatic principles in the
definition of the substabilities which are hierarchical and a part of a
larger organizational hierarchy.
19
Hierarchies are found useful in business practice in the areas of
both planning and control.
We find that nearly all problems of practical
interest, whether conceptual or operational, have to be partitioned and
hierarchically structured for solution.
Some of the major reasons for
Chester Barnard was among the first to do so in his Functions of the
Executive
.
19
If at any time we reach the undesirable hypothetical state where
the whole business organization is a huge homeostat both in the planning
and operational sense and such homeostat can be duplicated then we will
have reached stagnation.
17.
.
•
such partitioning are:
1.
20
Cognitive limitation of the intelligent organisms which
does not allow (a) operation on too many variables at a
time and (b) adequate specialization across all dimensions
of the problem.
The organism therefore must depend on a
substability for an appropriately preprocessed and timely
input.
2.
Intelligent organisms associate by transferring knowledge
from an area of expertise to a novel situation requiring
solution.
3.
In the process they become specialized.
Partitioned structures possess survival characteristics
which allow them to save many parts in case of failure.
Hierarchical arrangements allow the partitioned organization to
achieve an effective balance between centralization and decentralization
by considering issues such as:
1-
environmental and resource interdependencies between units
which argue for co-ordination or decoupling of their activities by a higher level;
2.
benefits from relative decentralization of subordinates such
as learning returns,
especially in an uncertain environment,
and motivational effects;
3.
benefits from relative centralization such as a reduction of
the number of transformations of objectives and thus of potential
20
For a more elaborate discussion of these issues and their implications for centralization, decentralization, standardization, and the
process of innovation, see among others: Schumpeter, Simon, Z. S. Zannetos
"
and 0. Poensgen, and Zannetos, "On the theorv
18.
of undue diversion of organization activities toward the
satisfaction of local objectives having little bearing on
underlying basic organizational objectives;
4.
communication requirements and costs necessary for the
coordination and control of units centralized from without
operating in an uncertain and changing environment
5.
;
and
capacities and costs of the individuals making transfor-
mations of objectives.
Perhaps because the developed mathematical tools begin to falter
when dealing with very large and complex systems, but more likely through
a lack of
appreciation for their importance, hierarchies of planning
and control have not been emphasized in some parts of cybernetic litera-
Hierarchies have been acknowledged as "speeding up" the action of
ture.
large, complex homeostats, as in the control of the firm
cently in the structuring of computers.
21
and more re-
Although we see ample evidence
of hierarchical structuring of software, we believe that the failure to
consider some of the fundamental notions underlying relatively decentralized hierarchical structures has contributed extensively to the present
rather sad state of software management.
Going further, we believe that hierarchies are essential to the
reduction of search spaces in most problems of practical interest.
21
Stafford Beer, Decision and Control, pp. 378-391.
19.
Hierarchical plans, incorporating hierarchical means-ends trees, can
reduce action spaces in a similar fashion.
22
Also, for reasons to be
indicated more fully later, hierarchies are useful in learning.
In
applying the strategy of simple reinforcement to a complex control
mechanism, we encounter a serious difficulty in distributing credit
for success of a complex strategy among the many components involved,
and thus inhibit causal diagnosis and learning.
This difficulty can
be managed by arranging for local reinforcement of partial goals within
a
hierarchy and by grading the training sequence of problems to parallel
a
process of organizational maturation.
Similarly, a hierarchy of per-
ceptual constructs reduces complex detail, implicit for example in our
vision, to a simplified version that higher-order cognitive processing
can handle.
C.
Control
It is in the area of control of operations that cybernetics exerts
a significant and undisputed influence.
Today so-called cybernetic con-
trols manage or help manage large chemical processes, provide automatic
quality control, facilitate traffic flows, schedule freight cars and
many other operating processes.
In this sense managerial functions have
been taken over, redefined, or upgraded with the aid of cybernetics.
The relatively sophisticated command and control systems used by the
military and by NASA whet our appetite for use of such techniques in
Newell and Simon, "GPS, A Program That Simulates Human Thought,
Computers and Thought
.
controlling not just operations, but the planning process itself.
For the latter, management could use help particularly in pattern
recognition and causal diagnosis.
23
Clearly one of the most obvious applications of cybernetics
to management is in the area of servomechanism control.
The business
organization's partial knowledge of its changing environment creates
many disturbances which must be corrected, some on a continuing basis.
Where relatively stable organizational resources exist for coping with
disturbances of relatively stable or predictable frequency distribution
(or probability distribution), straightforward analytical methods exist
for the working out of optimum policy parameter values.
24 25
'
Unfor-
tunately these methods tend to break down for complex non-linear control
situations
,
and one is driven to computer simulation
26
as an analytical
tool.
If one confines oneself to systems of moderate complexity one can
utilize the conceptual framework of servomechanism theory even where
its mathematical techniques are intractable.
One may usually
construct a simplified model of systems of even great complexity which
captures enough of the system's nature to be interesting and useful.
Further, the notion that all decisions may be viewed as embedded within
23
For some of the problems involved here see, Carroll and Zannetos,
"Toward Intelligent Management Information Systems."
Wiener, Cybernetics
25
H. A. Simon, "On the Application of Servomechanism Thoery in the
Study of Production Controls," Econometrica April, 1952.
,
Jay Forrester, Industrial Dynamics
,
1961.
21.
feedback loop systems is a powerful tool for exploring organizational
structure, even when the frequency spectra of the disturbances to be
controlled are too little known or too unstable to permit "optimal"
Fruitful design improvements may result from even very limited
design.
information, given an understanding of characteristic feedback system
behavior.
Often the incorporation of new channels of information into
important decision processes will markedly alter the dominant modes of
behavior of the system.
Unfortunately, in the area of cybernetics proper, there seems to
have been so far
an over- reliance on "black-box" input-output analysis;
this over-emphasis on "behavioristic controls" has so far limited the
applications of cybernetic principles in the area we feel the contribution can be of the greatest value.
The characteristic mode of analysis employed by Wiener utilized
as a complete description of a transducer its input-output relationship.
This "behavioristic" or "black-box" analysis is sometimes suggested for
use in practical management situations.
27
There are several reasons
why purely behavioristic analysis, as opposed to at least some structu-
ral-functional analysis of the parts of a transducer, may not be preferred.
In many practical situations one can realize an orders of magnitude
gain in efficiency if the knowledge gained in other situations is in-
corporated into the analysis.
Thus, for example, it may be that most
transducers of interest are constructed from parts whose behavior and
See Stafford Beer, Cybernetics and Management
.
structure is already knouTi.
Thus if we can efficiently identify such
parts in action within the transducer, we do not have to re-learn the
input-output relations for each part, and further, we can more efficiently analyze the input-output relationship for groups of parts to
determine the way in which these parts are related.
In addition, through
this functional analysis, we may gain knowledge about new parts which
can be generalized into new transducers.
Benefits such as those described above provide the main justification for using articulated models which can be used to fit more than one
Of course,
situation.
articulated models.
there are limitations on the efficient use of such
The transducer for instance may be too novel, or
too elemental, or the invariance of the input-output relations for the
whole transducer may be more assured than the combination of its parts
or, indeed,
of each part individually.
In general,
as we have already intimated, there will be a trade-off
which in most practical situations argues for a mixture of the two modes
of analysis.
Whenever pure behavioristic analysis is possible, one must
ask if such an approach is efficient.
In dealing with complex organiza-
tions as transducers, one often finds that many orders of magnitude separate the time required for a purely behavioristic approach from that re-
quired for practical results.
In addition, we have cases of practical interest in organizations
where pure behavioralistic analysis is not only less efficient but
impossible.
Let us suppose for instance, that the transducer undergoes
occasional change.
If the transducer is complex, one may not be able
to amass long enough sequences of input-output data to determine
ade-
quately the transformation applied by the transducer before its nature
is changed; but the sequences might well be long enough to determine
adequately the relationship through structural-functional analysis.
Of course, when we deal with cybernetic systems, as opposed to compu-
tational machines, the changing enviponment implied by conditions of
increasing entropy makes the observation of extremely long non-trivial
sequences very unlikely.
What we argue is for a proper balance between behavioristic and
structural analysis.
The needs of intelligent managerial systems re-
quire causal diagnosis and heavily favor functional systems.
In this
we feel that the study of complex real-world transducers with great
internal "memory""
must of necessity employ analysis of their parts.
Often, however, we may find that initially our only alternative is
behavioristic control, and we must therefore use it as means to the
higher end of more generalizable knowledge.
IV.
Some Concluding Remarks
In this paper we have attempted to analyze the strengths and weak-
nesses of the cybernetics approach as applied to management.
Our con-
clusion is that cybernetics can serve not only as a conceptual-philo-
sophical aid but also as an operational tool in both managerial planning
and control, but that so far, most of its promise is yet unrealized. This
^larvin Minsky, Computation
,
Finite and Infinite Machines
.
is especially true in the planning sphere.
Only in the area of control
of operations has the impact of this promising field shown tangible re-
sults
.
There is evidence, however, that the cybernetic sphere of influence is gradually expanding.
Soelberg
s
The work of H. Simon and his associates,
work in the area of sequential thought processes,
29
and
the work in the area of artificial intelligence software, some of which
is applicable for management, are typical examples of this expanding
influence.
Within the Managerial Information and Controls Group at the
Sloan School of Management at MIT we started a small project two years
ago in an effort to understand how people plan goal oriented search, and
how they sequentially adapt on the basis of experience in the context of
student learning.
The results of our experiments so far are encouraging.
Our ultimate goal is to develop educational systems which integrate basic
educational material, and guide the student toward the most efficient
search path once his goal is recognized.
But in order to reach our goal
we must first understand the process of transformation of goals to oper-
ation objectives, decisions, operations, etc. as described in our discussion of the management process.
In other words, we must learn how "trees"
of means to ends are generated, generalize on the basis of the common elements
of successful searches and then build systems which aid this planning proIn the broad sense, this is the same process which we must under-
cess.
stand if we hope to improve the management planning process.
29
Peer Soelberg, Conclusions from a Study of Decision Making,
Working Paper 173-66, Sloan School of Management, MIT.
25,
REFERENCES
Ashby, Ross, An Introduction to Cybernetics
,
New York, Wiley, 1956
Barnard, Chester, The Functions of the Executive
Harvard University Press, 1938.
Beer, Stafford, Cybernetics and Management
,
,
Cambridge, Mass.,
New York, Wiley, 1959.
Decision and Control: The Meaning of Operational
Research and Management Cybernetics London, New York, Wiley, 1966.
,
,
Brillouin, Leon, Scientific Uncertainty and Information
Academic Press, 1964.
New York,
,
Carroll, Donald and Zannetos, Zenon, "Toward the Realization of
Intelligent Management Information Systems," Proceedings of
Congress for Information Systems published by Thompson Book
Company, June 1967.
,
Cyert, Richard and March, James, A Behavioral Theory of the Firm
Englewood Cliffs, N. J., Prentice-Hall, 1963.
Minsky, Marvin, Computation: Finite and Infinite Machines
Cliffs, N. J., Prentice-Hall, 1967.
,
,
Englewood
"Steps Toward Artificial Intelligence," in Computers
and Thought edited by Edward Feigenbaum and Julian Feldman,
New York, McGraw-Hill, 1963.
,
,
March, James and Simon, Herbert, Organizations
,
New York, Wiley, 1958.
Newell, Allen and Simon, Herbert, "GPS, A Program That Simulates Human
Thought," in Computers and Thought edited by Edward Feigenbaum and
Julian Feldman, New York, McGraw-Hill, 1963.
,
Richardson, G. B.
"Imperfect Knowledge and Economic Efficiency,"
Oxford Economic Papers, N.S. Vol. 5, June 1953.
,
,
Samuel, A. L.
"Some Studies in Machine Learning Using the Game of
Checkers," in Computers and Thought edited by Edward Feigenbaum
and Julian Feldman, New York, McGraw-Hill, 1963.
,
,
Schumpeter, Joseph, The Theory of Economic Development
University Press, 1961 [c 1934].
,
New York, Oxford
26.
Shannon, Claude, The Mathematical Theory of Communication
University of Illinois Press, 1949.
Simon, Herbert, Administrative Behavior
Macmillan, 1965.
,
,
Urbana,
(2nd Edition), New YOrk,
Soelberg, Peer, "Conclusions from a Study of Decision Making," Working
Paper 173-66, Sloan School of Management, Massachusetts Institute
of Technology, 1966.
von Hayek, F. A., "Economics and Knowledge," Economica
February, 1937.
von Neumann, John, The Computer and the Brain
Press, 1958.
Wiener, Norbert, Cybernetics
,
,
,
N. S.
,
Vol.
A,
New Haven, Yale University
Cambridge, Mass., Technology Press, 1948.
Zannetos, Zenon S.
'On
he Theory of Divisional Structures: Some Aspects
of Centralization and Decentralization of Control and Decision Making,
Management Science Series B, Vol. 12, December 1965.
.
,
and Poensgen, Otto, "Innovation, Operational Control
and the Management Information System," The Business Quarterly
(Canada), Vol. 31, Winter 1966.
,
CD^
m
tASfc
6^
Mill *^-
6bl
"?DaO DD3 V7M
''"'«
IliltiJitllLlillJli,'!
TDfiD DD3 674 653
3
3
T060 003
3
TOfiD
fi7M
S^b
003 674 636
qm-G'f
3
T06D 003 674 612
^m-^'f
?060 003 674 650
Hl^ 6^
II
3
TD6D DD3 674 6D44
^-'''
lllllll
3
„
3
T06D DD3
„
&^°[
671
irM7irii'rr'll1'!"Mlllllll
1D6D 003 674 bEE
"'*
iiiifiiiiffi'r!iiii
3 TD60 003 TDl bb4
-(.^^
Download